stiga-huang commented on a change in pull request #934:
URL: https://github.com/apache/orc/pull/934#discussion_r727883073
##########
File path: c++/test/TestReader.cc
##########
@@ -101,4 +101,41 @@ namespace orc {
900, rowsInCurrentStripe, rowIndexStride, includedRowGroups));
}
+ void CheckFileWithSargs(const char* fileName, const char* softwareVersion) {
+ std::stringstream ss;
+ if(const char* example_dir = std::getenv("ORC_EXAMPLE_DIR")) {
+ ss << example_dir;
+ } else {
+ ss << "../../../examples";
+ }
+ // Read a file with bloom filters written by CPP writer in version 1.6.11.
+ ss << "/" << fileName;
+ ReaderOptions readerOpts;
+ std::unique_ptr<Reader> reader =
+ createReader(readLocalFile(ss.str().c_str()), readerOpts);
+ EXPECT_EQ(WriterId::ORC_CPP_WRITER, reader->getWriterId());
+ EXPECT_EQ(softwareVersion, reader->getSoftwareVersion());
+
+ // Create SearchArgument with a EQUALS predicate which can leverage the
bloom filters.
+ RowReaderOptions rowReaderOpts;
+ std::unique_ptr<SearchArgumentBuilder> sarg =
SearchArgumentFactory::newBuilder();
+ // Integer value 18000000000 has an inconsistent hash before the fix of
ORC-1024.
+ sarg->equals(1,
PredicateDataType::LONG,Literal(static_cast<int64_t>(18000000000L)));
+ std::unique_ptr<SearchArgument> final_sarg = sarg->build();
+ rowReaderOpts.searchArgument(std::move(final_sarg));
+ std::unique_ptr<RowReader> rowReader =
reader->createRowReader(rowReaderOpts);
+
+ // Make sure bad bloom filters won't affect the results.
+ std::unique_ptr<ColumnVectorBatch> batch =
+ rowReader->createRowBatch(1024);
+ EXPECT_TRUE(rowReader->next(*batch));
+ EXPECT_EQ(5, batch->numElements);
+ EXPECT_FALSE(rowReader->next(*batch));
+ }
+
+ TEST(TestRowReader, testSkipBadBloomFilters) {
+ CheckFileWithSargs("bad_bloom_filter_1.6.11.orc", "ORC C++ 1.6.11");
+ CheckFileWithSargs("bad_bloom_filter_1.6.0.orc", "ORC C++");
+ CheckFileWithSargs("bad_bloom_filter_1.6.11-SNAPSHOT.orc", "ORC C++
1.6.11-SNAPSHOT");
Review comment:
Yeah, this is for test coverage for the `SNAPSHOT` check.
These files are generated by the old codes without the fix. If we generate
them using the current codes, bloom filters in them won't have the issue. It
seems inevitable to introduce static files...
For `1.6.12-SNAPSHOT` vs `1.6.11-SNAPSHOT`, I think we will backport the fix
to 1.6 branch. The next release is 1.6.12. So 1.6.12 won't have the issue. Note
that we only skip versions of 1.6.0, 1.6.1, ..., 1.6.11, and 1.7.0. So I think
we should use `1.6.11-SNAPSHOT`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: dev-unsubscr...@orc.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
